222

Brain Research, 549 (1991) 222-229 © 1991 Elsevier Science Publishers B.V. 0006-8993/91/$03.50 ADONIS 0006899391166104

BRES 16610

The striatum and motor cortex in motor initiation and execution Erwin B. Montgomery Jr. and Steven R. Buchholz Department of Neurology and Neurological Surgery [Neurology], Washington University School of Medicine, St. Louis, M O (U. S.A.) (Accepted 11 December 1990)

Key words: Striatum; Basal ganglia; Motor initiation; Motor execution

The participation of striatal and motor cortex neurons in motor initiation and execution was studied using single neuronal recording in 3 monkeys performing wrist flexion and extension stimulus-initiated reaction time tasks. Observations of 46 striatal neurons whose activity correlated with the tasks were compared to recordings of 59 task-related motor cortex neurons. Neurons were classified as best related to the appearance of the go signal, movement onset, agonist or antagonist electromyographic changes, or the movement reaching target. Timing of neuronal activity changes in both striatum and motor cortex suggested that go signal-related neurons represent input function while most movement onset-related neurons represent output function. In the striatum, those related to reaching target represent output function. Furthermore, go signal-related neurons usually change activity before movement onset-related neurons which change activity prior to target attainment-related neurons. These observations suggest a hierachial organization within the striatum and motor cortex. Also the striatum participates in programming target acquisition as well as motor initiation.

INTRODUCTION A recent human study showed that m o t o r initiation may be distinct from execution 26. In that study, subjects made wrist extension movements to targets which changed location at different times during the task. Late target location changes resulted in two movements in which the first movement went to the original target location. However, changes in target location occurring approximately 100 ms after onset of agonist muscle activity could be compensated for and resulted in a single movement. These observations suggest that specification of trajectory, defining execution, does not occur until after m o t o r initiation. Other observations suggested that initiation and execution are mediated by different mechanisms within the striatum. Rostral caudate nucleus and putamen may be involved in initiation. Dopamine depletion of this region consequent to Parkinson's disease, may result in prolonged reaction times. Caudal caudate nucleus and putamen may be involved in execution and dopamine depletion of this region produces slow movement velocities (i.e. bradykinesia). Abnormalities of movement velocity programming relative to target conditions, demonstrated in human Parkinson's diseases subjects, suggest a basal ganglia role in programming motor execution as it relates to target acquisition 25. Therefore, it is likely that mechanisms exist

within the basal ganglia that are involved in programming target acquisition. Most previous studies of the striatum have concentrated on m o v e m e n t initiation 1-4'7-9. However, most basal ganglia neurons changed activity after motor cortex neurons, suggesting that putamen neurons are active too late to initiate m o v e m e n t 4's'9'19. Alternatively, others have suggested that the basal ganglia may be involved in scaling agonist electromyographic ( E M G ) activity 4'14'15. These inferences could be interpreted as suggesting a role in target acquisition. However, no prior study has directly examined target acquisition. Motor initiation and execution were studied in the striatum and motor cortex using single neuron recording in monkeys performing rapid wrist flexion and extension movements. The caudate nucleus and putamen are histologically identical and derive from the same embryonic structure. They are separated only by the development of the internal capsule. Therefore, data collected in caudate nucleus and putamen were pooled. A new analytic method was used which determines within a single task which behavioral event, such as the appearance of the go signal, movement onset, reaching target, or agonist or antagonist E M G changes, is best related to changes in neuronal activity. This method is an extension and quantification of a method introduced by Evarts 1° and is based on the premise that changes in neuronal activity are most consistently linked in time to

Correspondence: E.B. Montgomery Jr., Present address: Department of Neurology, The University of Arizona, Arizona Health Science Center, 1501 North Campbell Avenue, Tucson, AZ 85724, U.S.A.

223

the best related behavioral event. For example, neurons r e l a t e d t o t h e a p p e a r a n c e o f a g o signal w o u l d h a v e less v a r i a n c e in t h e l a t e n c i e s b e t w e e n c h a n g e s in n e u r o n a l a c t i v i t i e s a n d a p p e a r a n c e s o f a go signal t h a n to m o v e ment onset. Previous methods used carefully constructed behavioral tasks to identify physiological functions based on the consistency of responses across the different task c o n d i t i o n s . F o r e x a m p l e , a n e u r o n w h i c h c h a n g e s activity e a c h t i m e a specific m u s c l e is a c t i v e , r e g a r d l e s s o f t h e direction of movement

under different task conditions,

w o u l d b e classified as a m u s c l e - r e l a t e d n e u r o n 8"3°. H o w ever, studies have shown that the behaviors of many neurons are not necessarily consistent for different task c o n d i t i o n s 6'31. T h u s , it is n e c e s s a r y to u s e a m e t h o d w h i c h e s t a b l i s h e s b e h a v i o r a l c o r r e l a t i o n s w i t h i n a single task. Using this method, neurons within a structure can be s u b d i v i d e d i n t o d i f f e r e n t b e h a v i o r a l classes. T h i s allows c o m p a r i s o n s o f t h e o n s e t o f n e u r o n a l activities b e t w e e n c o r r e s p o n d i n g classes in d i f f e r e n t s t r u c t u r e s . T h u s , alt h o u g h g o s i g n a l - r e l a t e d n e u r o n a l a c t i v i t y in t h e m o t o r cortex may precede movement

onset-related neuronal

a c t i v i t y in s t r i a t u m , t h e s e s t r i a t a l n e u r o n s m a y p r e c e d e a c t i v i t y c h a n g e s in m o v e m e n t

o n s e t - r e l a t e d n e u r o n s in

motor cortex. This observation would support a hierarchical organization with respect to movement onset with striatum upstream of the motor cortex. Pooling data for each strucure would obscure the hierarchical organization.

tus and trained using only positive reinforcement. After training, a head holder and recording chamber were secured under sterile conditions using general anesthesias . Glass-insulated platinumiridium microelectrodes were inserted transdurally into the striatum and motor cortex. Extracellular action potentials of a single amplitude were counted in 10-ms bins. Wrist position and full wave rectified and integrated EMG of the wrist flexors and extensor muscles were sampled every 10 ms. EMGs were recorded using surface Ag/AgCI electrodes, 3 mm in diameter. The experiments were controlled by a 6502 microprocessor which also collected data for each trial. Data were transferred to a VAX computer system for off-line analysis. EMGs of the biceps, triceps, deltoids, shoulder muscles and paraspinal muscles were surveyed. In addition to the wrist prime movers only the biceps were consistently active with flexion tasks and were recorded with each neuron's activity. Neurons whose activity changes were related to biceps activity during flexion tasks were included as agonist muscle related. The monkeys were periodically given juice rewards independent of the behavioral tasks to assess possible neuronal responses to the juice reward. Neuronal responses found related to the juice rewards were excluded from further analysis. The analysis method used to assign behavioral classifications within a single trial is based on the variability of the monkeys' performance. Evarts TM demonstrated that when rasters of a motor cortex pyramidal tract neuron's activity were aligned according to the monkey's reaction time, the time the monkey initiated a response positively correlated with the time of neuronal activity change (Fig. 2). Thus, Evarts concluded that the neuronal response was temporally locked to wrist movement rather than to the go signal. The method used here extends Evarts' rationale by relating neuronal activity changes to multiple behavioral events. Thus, rasters aligned on the most relevant behavioral event have the most consistent alignment of neuronal activity. Figure 3 is a schematic representation of rasters for a hypothetical neuron whose activity changes are the most tightly linked to movement onset. Raster A is aligned on the go signal, and raster B aligns the same data on movement onset. Since the monkey's

MATERIALS AND METHODS

A Three Macaca nemestrina monkeys were trained to perform 70 degree stimulus-initiated, rapid (350 degrees/s) wrist flexion and extension movements. The monkey's hand fitted into a wedgeshaped handle which maintained finger extension (Fig. 1B). The range of motion was divided into 7 contiguous position segments represented by 7 target light emitting diodes (LEDs) arranged in a horizontal row (Fig. 1A). The internal segments were 12 degrees wide while the end segments were limited to 5 degrees by mechanical stops. Seven parallel cursor LEDs displayed hand position (Fig. 1A). All tasks were initiated by an end target LED illuminating (position 1 or 7 in Fig. 1B for extension and flexion, respectively) to which the monkey moved the handle to match the cursor to the illuminated target LED. After a hold time which randomly varied from 0.5 to 1.5 s, the target LED at the other end of the range of movement illuminated (7 for extension and 1 for flexion) signaling the monkey to move (go signal) within a maximum allowed reaction time of 250 ms. The monkey then held in the end target position for a random period of time ranging from 0.5 to 1.5 s after which a juice reward was given. Data from trials in which the monkey failed to initiate movement within 1 s of the appearance of the go signal were excluded from further analysis. Also, the monkeys had to move fully to the target (either position 1-7) within 500 ms in order for that trial's data to be analyzed. Thus, if the monkey paused during the movement the movement time would exceed the 500 ms limit and the data for that trial was rejected. Monkeys were gradually acclimated to the experimental appara-

Target LFOs =

i

0 0 0 0 0 0 0 I 2 3 4 5 6 @@®@®®@ !

7 I

Curso; LEDs

B

mecha rop

"~

~ss°s •St

I

. S~O

t

Fig. 1. The LED instruction panel (A) and the handle with the range of motion divided into 7 segments (B).

224

A

0.001 OOI-Z 0.05-i

R

. . . . . . . . . . .

•

.

. : ::'.'.'"

•

". ~

D ~ . . . .

0.001 x

i

~S" ~ , ~ . . _ . . . .

CLOt.

~s-.~_...~.,

~---t

500

0

..~...

i

, 500

i

. . . .

i

. . . .

500

F Neuronal Discharge I I IIIIIIIII

:

ll~

5OO

0

5O0

i

. . . .

i

. . . .

1

E

I

I

0

f -

i

500

Position

_1

11111111113

msec

•. : - : - . . . •

..:.:

.'.

;

• ,...:..., I I00

I

I 200

,~--

.----

:

...~

.

I 300

I

. . . . ....--..• I I I I00 200 300

-

Fig. 2. This figure from Evarts ]° shows rasters of neuronal activity aligned on the go signal (left margin). The raster on the left ordered the trials as they occurred. The raster on the right ordered the trials from the shortest reaction time (top) to the longest (bottom). Visual inspection allows the inference that the change in neuronal activity is better related to movement onset than to the go signal. (Reproduce with permission.)

reaction time is variable, the two rasters do not have the same structure. Spike counts for each trial within two adjacent time windows are compared in each of the two histograms. The distributions of the spike counts per trial within each window are shown in a, b, a" and b" (Fig. 3C,D). There is greater variability in a and b compared to a" and b'. A comparison of a and b using a nonparametric (Kolmogorov-Smirnov two-sample test) statistical method will show a lower chi-square value than a comparison of a" and b'. Thus, the P-value will be greater in the a and b comparison than in the a" and b" comparison. In the experiments, summed peri-event rasters and histograms of neuronal activity for the complete set of successful trials of a specific task were centered upon several behavioral events. These include the go signal, movement onset, achievement of target position, and onset of agonist and antagonist E M G activity. In the first method

50 mse¢ t I

A I

50 msec I

',

',

i

:

1

1

I

I

II

I'~,

I I I Illll

B

I ,IIIIIIII

II III I I I III

l

I

"go"

I

,t

lil illlll

I i

I li

i I

O

',11111111 I

I

I

tl

,

i i i i

i i

:a a alil II I!l ', I illlll',l

l

I ii

11 1 1 1 r, l l

',it ii i!ll i

illlll,,It I

I

II

b

movement onset

signal

D

C no. of trials j

,

• i

i

•

b'

i .

0 2 4 0 2 4

L

O'

!

NO. OF SPIKES PER TRIAL

I

NO. OF SPIKES PER TRI=AL

Fig. 3. Rasters of neuronal activity (spikes) of a hypothetical neuron aligned on the go signal (A) and movement onset (B). Two adjacent windows are examined (a and b and a" and b'). The distribution of spikes per trial for each window is shown in C and D.

I

I

I IIIIIII

. J /

Fig. 4. Schematic demonstration of the methods used to determine which behavioral event is best related to changes in neuronal activity based on histograms centered (time 0) on various events (C) and trial-by-trial analysis (F). In the first method (degree of change), a time period represented by the open box (100 ms) is compared to the adjacent shaded box (100 ms) using the Kolmogorov-Smirnov two-sample test (D). Both boxes are moved through time, one bin at a time and the chi-square value determined and plotted in A. When the plot reaches the 0.05 level the P-value is 0.05, and similarly for the 0.01 and 0.001 levels. The second test of relatedness (change from baseline) uses a similar method except that the open box (100 ms) is compared to the shaded box (500 ms) which represents baseline activity prior to the go signal (E and B). The third test (inter-trial variance) uses a method which identifies homologous regions in the trains of neuronal activity for each individual trial and measures the latencies to various behavioral events (F).

(degree of change) the summed spike counts across multiple trials in a 100-ms window (10 bins) were compared to spike counts in an adjacent 100-ms window (10 bins) (Fig. 4D). The two 100-ms windows were moved through time at 10-ms intervals over a period from 500 ms before to 500 ms after the time of the behavioral event• The chi-square value was plotted (Fig. 4A). When the graph of the chi-square value reached the 0.05 level, the neuronal activities within the two windows were significantly different to a P-value of 0.05; similarly for the 0.01 and the 0.001 levels• The second method (change from baseline) is similar except that neuronal activity within the 100-ms window was compared to neuronal activity within the 500-ms window during a hold period which preceded the go signal in each task (Fig. 4E). The results are shown in Fig. 4B. An example of a go signal-related neuron is seen in Fig. 5. Rasters, histograms, and the associated plots of the chi-squarc values are centered on the go signal (Fig. 5A), movement onset (Fig. 5B) and final target attainment (Fig. 5C). Approximately 140 ms after the go signal there is a brief increase in activity followed by a sudden inhibition (point a in Fig. 5). The plot of the associated chi-square statistic for the inhibition (point a" in Fig. 5) reaches the lowest P-value for the histogram centered on the go signal; thus, the inhibitory change is best related to the go signal. The third method (inter-trial variance) identifies homologous regions in the trains of neuronal activity for each individual trial and measures their latencies to behavioral events (Fig. 4F). It is assumed that the set of latencies with the least variance indicates the event to which the neuronal activity is most tightly linked. This method has been described in detail previously 24. In practice, the results of the three methods were generally consistent• Occasionally, the P-value of the first method showed a difference between rasters centered on different behavioral events while the second method reached the minimum P-value for all rasters. The first method would allow a distinct assignment of behavioral type while the second method would not. In this case, the first method was used for behavioral classification. The third

225

o' I;

A

b

0001 .. GOI~ i

.

.

.

"

:•

I

.

~001..' GOI~ 0.05"

1,

,

';

'

I

_

'

I I

l II I

I

o

,' ,."'

I

II

""'

II II I I

I

I

II I

.

I I

' l'

b I

I

I

I

I

I , I I ,I

I I

,o,..,,

,

'y,,,,,

.i,

.,*l,,w

I

,i,__ i,.,*

I

I .I

•I

• ,I i s

* I I , ii*l ,

, III ; , ;I I IIM

I, , . I • I II

I I II • I II

......

I I

I IIIIlllll

l

I

,,. •

i I III

I

I

, ,,

I'll

I I II I II i I I I l i l l l l l

o

E,

cob' 1 l "

GO01-.. O.Ol~ GO01,,. .

.

.

.

.

.

.

.

.

b .

-,',,

.

(lol-;

I

~ t ( ~ ,' , .

.

..~ y ~ , T - - ~ : - - . J

500

, I ;"q'

..

I

I

*1 0

500

I

I

,

: II

'

I

I% I I I II I I

I

Ill

Ill

II

iii

I I

I

•

0

I

,

I

I

I

{

II

I

I,,,

",,,,

, '",,,.,,,' '", I

'I l l ", : ,'; , I', I 1',' ' , ' . :,

I

II

":, ',', ,~ ','; ,=,.,

I

I

•

; I,I

,',,'

Ul I I

li

II

I

,:

II

I

' ' ° "'I I"l l"l

'

_,."'",=.,,,=,;,;,

..SO0

b'

C

aool.. 0.01~

*'

:

b

l

~ T

A

Q

* I II

0.(X)I~, (101 -

I i

.

.

,

.

.

i

b,

•

•

•

•

o

po

"::''

1 I

I III II II i

i

.

I Ill I II II I •

II

lU I

• II

II

III I III III II I I I I I I • I tl I l l

5OO

0

5OO

5OO

I I

I I

I I

I Ill III I IIIIII I I I II | I I I I III I II l l l l i l l l l l l I

• I II

0

! I

I I

li I

I

, '',,

I

II

I1: I

! • I

l I

I I I

I I II I III llillllll II I I II II I I II I III IIII I l l III I IIII I l l l l • I llllll I I I I IIIIII II

5OO

Fig. 5. Histograms, rasters, and plots of statistical measures for a go signal-related neuron centered (time 0) in (A) the go signal, (B) movement onset, and (C) final target attainment. The baseline activity is tonic; after which there is a sharp increase of 140 ms after the go signal. This is followed by inhibition best related to the go signal (point a) as evidenced by point a" in the plot of the chi-square value as determined by the method described in Fig. 4D. The neuronal activity returns in a bursting fashion which is best synchronized to the time of final target attainment (point b) evidenced by point b" in the chi-square plots.

method (inter-trial variance) was used when the plots of the chi-square value in both the first and second method for all rasters reached minimum P-value. Some neurons might be encountered for which all histograms reached the minimum P-value and the latency variabilities for the third method were at the minimum for more than one behavioral event. The behavioral correlations for these neurons would be judged indeterminant for that task.

RESULTS

In 3 monkeys, 136 neurons were recorded in the striatum of which 46 significantly changed activity relative to at least one task. The anatomical locations of the related neurons are shown in Fig, 6. In motor cortex, 150 neurons were recorded of which 59 changed activity related to at least one of the tasks. The anatomical

location of electrode penetrations containing the recording sites for active neurons are shown in Fig. 7. Table I shows the classification of these neurons according to their best behavioral correlation. In the striatum, 21 neurons (45%) changed activity with both flexion and extension, while in motor cortex 18 (35%) were bidirectional. Of the bidirectional neurons in the striatum, 2 were not consistent in their behavioral correlations for both tasks, while 2 in the motor cortex were not consistent. In 3 striatal neurons, there were serial changes in neuronal activity in which the first change was best related to the go signal, but subsequent changes were best related to final target attainment. For example, one neuron's tonic activity seen prior to the go signal changed 140 ms later to a sharp burst followed by

226

'"i~'~"

r\\Pi

"

.

.,-,

o'.

Fig. 6. Anatomical distributions of neurons related to the tasks for each monkey (A,B,C). Each figure represents a series of sequential sections taken through the basal ganglia. Sections are separated by 1 ram. O, neurons related to the go signal; Q, movement onset-related neurons; A, final target-related neurons; A, EMG onset-related neurons.

A

B 2O

inhibition (Fig. 5, point a) best related to the go signal. Activity returned (Fig. 5, point b) and became periodic which synchronized best when the raster and histogram

mm

15

10

i

i

5 i

central sulcu$

0 5 10 mm

TABLE I

J

Behavioral classifications o f initial neuronal activity changes which had a significant change in their activity related to either task Since there were two tasks, a neuron could have a significant change in activity with both tasks and data from both tasks would he included in the analysis. Thus, the number of data points could exceed the number of neurons analyzed. In the striatum, 21 neurons were active in both tasks. In the motor cortex, 18 neurons were active in both tasks.

C 20

D

mm

,,5

to

5

o

2O

Striatum Motor cortex

Movement onset

Reaching target

Agonist EMG

Antagonist EMG

15 33

37 28

19 7

2 5

0 4

mm

15

10

i

i

5 i

central sulcus

As

0 5 mrn

I0

'Go' signal

15

anterior

/

10

/

15

Fig. 7. Scheme of the distribution of electrode penetrations containing neuronal recording sites in the motor cortex (A) for each monkey (B,C, and D).

227 MOTOR CORTEX "GO" SIGNAL RELATED NEURONS

o

,, ,,,I, III I,, 360 26o ibo ~o

,60 260 36o

MOTOR CORTEX MOVEMENT ONSET REtATED NEURONS

'1

o

Itl

I

t ,,11

a6o 200 ~oo ~o

~6o 260

3bo

AGONIST EMG

'1

o

I

, ,I,h],,L

360 200

,6o ~o

i6o ~o

3bo

TR,A AL'GO" ,GNAL RELATED NEURONS I I

0

III

I I

II

I

36o 2b0 160 ~o ,60 260 360 STRIATAL MOVEMENT ONSET RELATED NEURONS

o

I II, .I,II,

,,, ,,.,

3bo 2bo 16o ~o

,6o 200 s6o

51"RIATAL REACHING TARGET RELATED NEURONS

,I, .I,.,

,,,.

23 36o '2~o ,6o ~o ,6o o

ill

I ,,,,

3oo 2~

,ill

,60

,

~T

I

I

26o 360

i

160 2